Application of hydrogen direct injection in the cooperative fuel research engine and optimization of standardization control under lean conditions

Hydrogen is a promising zero‑carbon fuel for internal combustion engines (ICEs), while Cooperative Fuel Research (CFR) engines provide a standard platform for knock related fuel evaluation due to their variable compression ratio (CR). In this study, hydrogen direct injection (DI) was implemented on a CFR engine to establish a stable lean-operation reference for future hydrogen octane-rating and knock studies under DI conditions. Pilot tests were conducted over a CR range from 6.5 to 14, to examine the effects of excess-air ratio (λ), start of injection (SOI), and combustion phasing, represented by the crank angle at 50% mass fraction burned (CA50). The results show that indicated thermal efficiency (ITE) first increased with CR but decreased clearly when CR exceeded 12. Under similar CA50 and λ conditions, increasing CR from 12 to 14 reduced ITE by more than 3 percentage points. GT-Power analysis indicated that blow-by through the side mounted spark plug port at CR = 14 could account for more than 2.5 percentage points of this efficiency loss, while restricted early flame development near the spark plug region may further contribute to the performance decline. At high CR, near-stoichiometric mixtures tend to induce preignition and knock, leading to increased variability. The effect of SOI was weaker and less monotonic than those of CR, λ, and CA50. Nevertheless, retarding SOI generally improved ITE. At CR = 12, ITE increased from 29.9% to 33.0% as SOI was delayed from −140 to −50 crank angle degrees after top dead center (°aTDC). CONVERGE-CFD simulations showed that delayed injection promoted a favorable stratified mixture near the spark plug, accelerating early combustion. Response surface models for ITE, coefficient of variation of net indicated mean effective pressure (COV_IMEPn), and nitrogen oxides (NOx) were then developed using a four-factor, three-level Box–Behnken design. The simplified ITE model showed the best predictive performance, with R² > 90% and a predicted R² of 89.69%, whereas the COV_IMEPn model showed weaker predictive capability. Desirability based optimization predicted the standardized operating condition. Validation tests yield ITE = 34.57%, COV_IMEPn = 1.35%, and NOx = 10.2 ppm, with good agreement with predictions and an efficiency gain of about 1.5 percentage points. Overall, by testing and analyzing hydrogen DI over a wide range of operating conditions, this study provides a practical basis for standardizing hydrogen tests on the CFR platform and lays the groundwork for future research into hydrogen fuel knock evaluation under DI conditions.